Interarboreal Operations: Head Movement and the Extension Requirement

Author(s): Jonathan David Bobaljik and Samuel Brown

Source: Linguistic Inquiry, Vol. 28, No. 2 (Spring, 1997), pp. 345-356
Published by: The MIT Press
Stable URL: http://www.jstor.org/stable/4178980 .
Accessed: 14/02/2014 09:49

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to Linguistic Inquiry.

http://www.jstor.org

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions

INTERARBOREAL OPERATIONS:
HEADMOVEMENT ANDTHE
EXTENSION REQUIREMENT
Jonathan David Bobaljik
Harvard University
Samuel Brown
Harvard Medical School
This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM
All use subject to JSTOR Terms and Conditions
Squibs
and
Discussion
Chomsky(1993) suggests a returnto the "bottom-up"structure-build-
ing algorithmof a generalized transformation(GT), applicationsof
which are restrictedby the Extension Requirement(ER) discussed
below. Chomsky notes that two central processes-head movement
via adjunction(or adjunctiongenerally) and LF movement-appear
to violate the ER. Recent work has attemptedto derive the exceptional
natureof these two processesfromindependentprinciplesof grammar.
We contend that the seemingly exceptionalnatureof head movement
results from an unmotivatedand unrecognizedassumptionthat effec-
tively prohibits what we will call interarboreal operations. If this
assumption is abandoned,head movement can be seen to obey the
ER. In this squib we therefore offer a formal solution to a theory-
internalproblem, in the hopes of suggesting a direction for further
research.
An example of an interarborealoperationis schematizedin (1).
In this derivation(which we discuss in more detail in section 2), the
verb adjoins to I? prior to concatenationof I? and the VP by the
operation Merge (a subcase of GT), as in (lb). It is the resulting
complex I? that is subsequentlymerged with the VP, projectingan IP
node. As a result, the phrase marker generated by this opera-
tion-(lc)-is isomorphicto thatderivedby the more traditionalER-
violating version of head movement. Within the currentframework,
thereis no principlethatexcludes operationslike (1). Thus, we suggest
that either such interarborealoperationsshould be admittedand ex-
ploited, or a principledreason to exclude them should be proposed.
Here, we will take the formertack.
In section 1 we present Chomsky's definition of the ER and
examine the apparentlyexceptionalnatureof head movement.In sec-
tion 2 we clarifythe technicaldetailsof the ER-consistentinterarboreal
operation that derives head adjunctionconfigurations.In section 3
This squib reportson ideas developed independentlyin Bobaljik 1995a
and Brown 1995. For discussion of these ideas and comments on the squib,
we would like to thankNoam Chomsky,Samuel D. Epstein,Danny Fox, How-
ard Lasnik, Susi Wurmbrand,and two LI reviewers. Errorsare ours alone.
345
346 SQUIBS AND DISCUSSION

(1) a. b. c.
VP v IP

V DP I V I VP

I V V DP

we discuss how overgenerationis preventedby integratinga slightly

modified Chain Condition(Rizzi 1986, 1990), similar to Chomsky's
(1995b:253) C-CommandConditionon chains.In section4 we present
our conclusions.

1 The Extension Requirement and the Problem of Head

Movement
Chomsky (1993:22) suggests that the basic syntactic operationis a
generalized transformation(GT), which may be broken down into
discretesteps as in (2) (afterChomsky 1993:22, Kitahara1994, 1995).
(2) a. Target a phrasemarkerK.
b. Add an empty 0 externalto (i.e., as a sister to) K.
c. Substitutea phrasemarkera for 0, formingthe new phrase
marker y = a projection of K.
In subsequentwork (Chomsky 1995a,b) the operationin (2) is called
Merge, thoughthe steps in (2) remainessentiallyunchanged.Move (or
Move a) is a similaroperation(Chomsky 1993:22),with the additional
assumptionsthat,Bin (2c) is an element(nonrootphrasemarker)within
the target K and that the operationinvolves a copying and/or chain
formationprocedure.' Chomsky also proposes (1993:22) that GT is
constrainedby what has come to be called the ExtensionRequirement,
(3).
(3) "0 must be external to the targetedphrase markerK [i.e.,
K]. Thus, GT and Move a extend K [i.e., K] to K* [i.e., -y],
which includes K [i.e., K] as a properpart."

1 The precise details of the copying procedureare differentin Chomsky

1993 and subsequentwork. In Chomsky 1993 it is assumed that movement
"leaves behind a trace" (p. 22) identicalto the moved element (p. 37). In the
text we adopt the formulationof Move given in Chomsky 1995a:fn. 13.
(i) Move a
"[G]iven the phrasemarker ;, select K, a s.t. the root of K dominates
or c-commandsa; form y by merging Ot, K (what we call "targeting
K and raising a"); define the chain CH = (aLI, l2) relationallyas the
pair (a, at) . . ."
For Chomsky's (1995a) purposes, the choice between this formulation
and one in which the chain is formed as a part of movement (p. 399) is not
important.For our purposes, it becomes importantin section 3.

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions
 SQUIBS AND DISCUSSION 347

In other words, the ER requiresthat all operationsadd a sister to a

root node. A new sister node cannot be created for any node that is
alreadyin a sisterhoodrelationship.This blocks downwardand side-
ways movement and other noncyclic operations.However, Chomsky
(by stipulation)restricts(3) to overt substitutionoperationsonly. The
reason for this is clear: adjunction(especially head movement) and
(most) LF movement violate (3). To see this, consider (4), a typical
case of V-to-I raising.

(4) IP

I VP

V
NPsubj

V NPobj

Under standardassumptions,the head V cannot raise to I? until I? is

in the same phrase markeras the V-that is, until after mergerof Io
and VP creates the I' or IP node. However, verb raising after I? has
been introducedandprojectedas IP does not extendthe rootnode-IP.
Thus, the operationviolates the ER.
Kitahara(1995) andWatanabe(1995) attemptin differentways to
derivethe apparentlyexceptionalnatureof headmovement,adjunction
generally, and LF movement from principles of economy.2 We will
show that head movement is not exceptional if one abandonsthe as-
sumptionthatthe element thatmoves must originatewithin the target,
an assumptionfor which no independentmotivationis provided.3'4

2 Frankand Kroch (1995) propose to eliminate the ER entirely, deriving

strict cyclicity effects with the formal apparatusof Tree-AdjoiningGrammar
(TAG). They do not specifically addresshead movement in this context, and
a comparisonof the formalmachineriesof TAG and the theoryproposedhere,
althoughprobablyultimatelyinstructive,would take us far afield.
The questions addressedhere arise in a quite differentmannerunder the
"AttractF(eature)" hypothesis put forth in Chomsky 1995b:297ff., and even
more so given the modificationsof that approachsuggested in Chomsky's fall
1995 class lectures,in particularthe proposalthatfeaturesmay move indepen-
dently of the terminalnodes (Xos) under which they are organized.Although
this allows for a differentinterpretationof, andperhapssolutionto, the problems
addressedhere, the relevant proposals in Chomsky's fall 1995 class lectures
representa nontrivialdeparturefrom Chomsky 1993, 1995b.
3We will not discuss LF movement here. An alternativeformulationof
LF movement that does not violate the ER is made available by adopting a
copy theory of movement (Chomsky 1993). Given that a chain consists of
multiple instances or copies of a single element, each occupying a distinct

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions
348 SQUIBS AND DISCUSSION

Consider again GT, (2). This can be viewed as the core of the
operationsMerge and Move (e.g., Chomsky 1993:22, Epstein 1994).
In either case, it is assumed that the target (K in (2)) must be a root
node-a phrasemarkerthatis not a subtreeof any otherphrasemarker.
In the case of Merge (or binaryGT; Chomsky 1993:22), the element
that is substituted(I in (2)) is, like K, a root node; it may be either
an independentlyconstructed,arbitrarilycomplex phrasemarkeror a
term selected from the numeration.It is importantto note that, within
this framework,multiple phrase markersoften must be built up in
parallel,thoughthis point is frequentlyomittedas an expositoryshort-
hand.The operationMerge concatenatesthese phrasemarkers,apply-
ing "often enough to leave ... just a single [phrasemarker]"(Chom-
sky 1995b:226, 243). To see that there must be multiple (complex)

structuralposition, "overt movement" may be characterizedas the pronuncia-

tion of the highestinstanceor copy, with "covert (i.e., LF) movement" pronun-
ciation of the lower copy. Metaphoricallyspeaking, all movement could be
overt, the apparentovert/LFdistinctionbeing an artifactof pronunciation.Ap-
proaches along these lines are pursuedin Bobaljik 1995b, Brody 1995, Groat
and O'Neil 1994, Pesetsky, in preparation,and works referredto therein.
4 Chomsky (1993) and Kitahara(1995) consider two types of (overt)

adjunctionthat apparentlyviolate the ER. One of these is head movement,

which we consider in detail in the text. The other is an asymmetrybetween
relativeclauses and complementclauses concerningConditionC "reconstruc-
tion" effects. Thus,given a pairlike (i) and(ii) (fromFreidin1994), coreference
betweenJohn andhe is possible if John is in a relativeclause (i), but impossible
if John is in a CP complementto the noun recommendation(ii).
(i) [Which recommendationthat Johni revised] did hei/k
submit tDp?
(ii) [Which recommendationthat John, should be promoted]did hei/k
submit tDp?
As a reviewer points out, an appeal to interarborealmovement along the
lines we suggest for head movement does not seem to be a plausible route to
an account of the contrastbetween (i) and (ii). However, as Chomsky (1993:
n. 44) notes, there are a numberof alternativeaccounts of (i) and (ii) that do
not appealto countercyclicadjunction.Freidin(1994:1383-1384) suggests that
in (i), "the relative clause modifies the variable and thereforeremains as a
part of the quantifierstructure,as in [(iii)]"; coindexation does not lead to a
ConditionC violation. In the (abbreviated)LF representationof (ii) shown in
(iv), the complementclause is a part of the variable,not a modifier of it, and
coindexationthereforeviolates ConditionC.
(iii) (for which x such that John revised x, x = a recommendation)did
[lP he submitx]
(iv) (for which x, x = a recommendationthat John should be promoted)
did [Ip he submitx]
In other words, Freidinproposes that only the variableneed reconstruct,
andthata relativeclause, as a modifierof the variable,need not. The asymmetry
is capturedwithout appealto countercyclicadjunction,though otherquestions
are raised.It is worthnoting thatChomsky(1993:n. 44) cites Freidin'sanalysis
as evidence that binding conditionshold at LF. Otheraccounts of the contrast
are possible, though such discussion is well beyond the scope of a squib.

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions
 SQUIBS AND DISCUSSION 349

root nodes at various stages of the derivation,consider the inputs to

an operationof Mergethatintroducesa complex specifier,for instance,
a "VP-internalsubject." This is illustratedin (5).

(5) Merger of two independentroot nodes

DP V'

D N(P) + V DP =
the monster eat

D N(P)
the pizza
VP

DP VI

D N(P) V DP
the monster eat

D N(P)
the pizza

By virtue of the ER, separateoperationsof Merge must constructthe

(subject) DP and the V' in parallel. It is neither the D [the] nor the
N(P) [monster]that merges with the V' to create the specifier of VP,
but the independentlyconstructedcomplex (root) node, DP. For the
same reason, if both the subjectand object containedrelative clauses,
then Move operations would have to occur internal to each phrase
markeras well, priorto the operationof Mergethatconjoinsthe subject
DP and the V'.5 In sum: the ER requiresthat phrasemarkersbe built
up in parallel, and thus the derivationmust consist-at least in the
early stages-of multiple root nodes concatenatedby the operation
Merge (see Chomsky 1993:22).
The operationMove differs from Merge in that, for Move, the
element that is merged with the target K is not an independentroot
node, but a copy of an element already within some existing phrase
marker.Move thusinvolves an extrastep of copying:6the suboperation
Copy takes some element in a phrasemarkerand creates an identical

5 As a reviewer correctlypoints out, the ER prohibitsmovement internal

to a phrase marker.What is meant here by each phrase markeris the output
of Merge, that is, each y in (2c). The point is simply that the grammarmust
allow mergerof phrasemarkersseparatelyconstructedby applicationsof Move.
6 As Noam Chomsky (personal communication)points out, it could be
that both Move and Merge involve the suboperationCopy, whereas Merge,
not Move, requiresan extra step, Delete.

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions
350 SQUIBS AND DISCUSSION

copy. This copy then serves as the input (3) to the substitutionstep
of GT (2c). We illustratethis schematicallyin (6), where a DP is raised
from complement of V to specifier of IP. First, Copy operates (6a);
then GT/Move substitutesthe copy (6b).

(6) a. IP

D + I VP

thD r N(P)V D
\the arrieves/
monster/

the monsr m

b. IP

DP + I VP

themonster V DP
arrives

the monster
IP

DP I'

the monster IVP

V DP
arrives ...........

the monster

At PF the lower instanceof the DP [the monster]is deleted (Chomsky

1993:35).
As noted above, Chomsky makes another assumption about
Move, namely, that the source of the copy must be contained in the
target (see 1993:22 and 1995a:399 fn. 13; also see fn. 1 above). Ob-

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions
 SQUIBS AND DISCUSSION 351

viously, this assumptionprecludes interarborealoperationslike (1);

however, we are awareof no independentmotivationfor it. Moreover,
though the assumptionis tacitly maintainedin Chomsky 1995b, there
it is apparentlytakento follow fromthe requirementthatevery instance
of a given element must either c-commandor be c-commandedby all
otherinstancesof thatelement (see Chomsky 1995b:250,253). In fact,
as we will show in section 3, this is not correct. The C-Command
Condition (Rizzi's (1986) Chain Condition) does not itself rule out
interarborealoperations;they are excluded by stipulationonly. We
now consider a potential advantageto abandoningthis stipulation.

2 Interarboreal Operations
Consideragain the derivationschematizedin (1). The phrasemarker
in (la) is the result of the following operation:GT targetsa root node
K (= V?) and adds an empty position 0, projecting y (= VP). In this
instance, the element substituted for 0 is DP (independentlycon-
structedby GT if complex, or providedfrom the numerationby Select
if trivial/lexical).The next step derives the phrasemarkerin (lb): GT
targetsa root node K (= 10) and adds an empty position 0 projecting
,y ( = complex 10).As in the first operation,the targetis a trivialphrase
marker-in this case lo-provided from the numerationby Select.
We take this step to be adjunction;hence, -y is a segment of 10 (i.e.,
it does not projectto a higherbar level).7 In this step the element that
is substitutedfor 0 is V?-a copy of an element containedin a more
complex phrasemarker.That is, this operationis Move (i.e., Copy +
Merge); it obeys the ER and all aspects of the characterizationof
movementexcept for the unmotivatedstipulationthatMove mustoper-
ate internallyto a single phrasemarker.Finally, yet anotheroperation
of GT targetsa root node K (the complex 10from (lb)) and substitutes
the VP for 0 projecting y (= IP). All three operationsobey the ER.
Importantly,the result of the final step (1c) is exactly the structure
that is derived via the ER-violatingmovement operationin (4)-the
standardcharacterizationof head movement. The problem, then, is
not to derive the exceptional natureof head movement, but to show
that interarborealoperationscan be adequatelyconstrained.

3 Constraining Interarboreal Operations

Many questions arise at this point, and we will addressa few of them
here. First, we consider the status of XP-movement with respect to
interarborealoperations.We also consider the status of requirements
such as ShortestMovement-in particular,the Head MovementCon-
straint(Travis 1984). We argue that Rizzi's (1986) Chain Condition,

7This follows fromthe assumptionthat " [a]djunctiondiffersfrom substi-

tution ... only in that it forms a two-segment category rather than a new
category" (Chomsky 1995a:402;see also May 1985 and,for potentialproblems
with this view, Epstein 1989).

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions
352 SQUIBS AND DISCUSSION

or some similarrequirement,will bothexclude interarborealXP-move-

ment (at least via substitution)and permitMinimality/ShortestMove-
ment effects to be captured.Note also that we assume, with Chomsky
(1993:22, 1995b:226, 243), that the derivationconverges only if suc-
cessive applications of Merge derive a single phrase marker at or
before LF.8
Rizzi (1986:66, 1990) proposes essentially the following condi-
tion on chains, thatis, on multiple instancesof a single term;compare
Chomsky's Shortest Movement and Minimal Link Condition (e.g.,
1995b:296):
(7) Chain Condition
CHAIN = {al1, . . ., o, } such that ...
a. ot, c-commandsan,+1
b. 7 3 1 such that an, c-commands ,B,1 c-commands t, +1
where 1 is of some relevant type (e.g., same features as ox)
[i.e., Relativized Minimality]
Rizzi proposes (7) as a condition on the well-formednessof the
final (LF) representation.We insteadfollow Chomsky(1995b:253) in
taking it to be part of the definition of Merge. The output of every
application of Merge is evaluated with respect to (7) (and perhaps
other conditions on the operation).9Thus, we take this evaluationto
be derivational-it is local in the sense that it is evaluated at every
step of the derivation,without referenceto the effects of subsequent
operations.10In any phrase markerderived by Merge containingtwo

8
This version of Inclusiveness is defended and extended in Bobaljik
1995a. In particular,interarborealoperationslogically admitthe possibilitythat
a copy of an element could check features in a phrase markerthat is never
containedin the final phrasemarker.Inclusiveness excludes that possibility.
9 As Chomsky writes, "A chain CH = (cx,t(ca))formed by Move meets
several conditions, which we take to be partof the definition of the operation
itself. One of these is the C-Command Condition: a must c-command its
trace.. ." (1995b:253).
We differ from Chomsky solely in our interpretationof (7a). Chomsky
does not consider interarborealoperations,whereas we assume that (7) holds
only internalto the phrase markercreated by the operationMerge (i.e., y in
(2))-presumably, c-command is undefinedacross phrase markers.
Interarborealoperationspose problemsfor the definitionof locality in both
a "Shortest Movement" (Chomsky 1993) and an "AttractClosest" approach
(Chomsky 1995b); neither will suffice to constrain interarborealoperations
given the problemof defining closest or shortestacrossphrasemarkers.Invok-
ing (7b) as a constrainton chain formationvia applicationof Mergepotentially
avoids these problems-though, as a reviewer points out, unlike Chomsky's
C-CommandCondition (7a), relevant elements for (7b) not only include the
element "moved" and its trace/copy,but also potentiallyinvolve other copies
of this element more deeply embedded in a given phrase marker.(7b) may
requirea weaker notion of locality than Chomsky's condition does. We leave
the matteropen.
10This follows, for example, the definition of local in Collins 1995:72:
"At any step S of the derivationD, the decision about whetheror not to apply
an operation0 is made on the basis of informationavailable at S." This con-

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions
 SQUIBS AND DISCUSSION 353

instances of some term ot, the two instancesare interpretedas a chain

iff (7) is satisfied. Taking, for example, the interarborealoperationin
(1), an operation of Merge creates the phrase markerin (lb). This
operation satisfies (7) since the newly created phrase marker-the
complex Io in (lb)-contains only one V. A subsequentoperationof
Merge joins the two phrasemarkers(la) and (Ib). The newly created
phrasemarker(lc) containstwo Vos;thus, (7) is satisfied-the opera-
tion of Merge is legitimate-only if these are in the appropriaterela-
tionship. As noted above, underthe segmentationanalysis of adjunc-
tion (May 1985) the required c-command configuration obtains if
interarborealhead movement is analyzed as adjunction.In addition,
Baker(1988) providesevidence thatheads c-commandout of adjunc-
tion structures;that is, a verb adjoined to I governs its trace. Thus,
we take interarborealhead movement to be adjunction.The Chain
Condition, as part of the definition of Merge, is satisfied.
Turningnow to XP-movement,consider a schematicexample of
an interarborealoperationof XP-substitution.1' In (8) the complement
DP of some head X moves to become the specifier of an independent
(root) phrasemarker,YP (i.e., the DP is copied and the copy merges
with Y'), satisfyingthe ER. Y' projectsto YP, andYP in turnis merged
with the root node containingthe source element DP, again satisfying
the ER. The result looks as if "sideways" movement had occurred.
(8) clearly runs afoul of the Chain Condition. The final operation
derives a phrasemarkercontainingmultiple instances of a single DP
(i.e., the original as complementto X and a copy as the specifier of

trasts with "global" evaluation procedures,which may "look ahead" and

considersubsequentsteps of the derivationin determiningeconomy. Formally,
it is similarto Lasnikand Saito's (1992:90) GeneralizedProperBinding Condi-
tion-the principle that traces must be bound must hold at every step of the
derivation.As a reviewer notes, to the extent that any (nonbare)output filter
can be restated as such an "everywhere" condition, the derivationalversus
representationaldistinctionis blurred,andthe case for a "strictly derivational"
approachweakened.
l As David Pesetsky (personalcommunication)points out, many hypo-
theticalderivationsinvolving interarborealXP-movementsuperficiallyresem-
bling (1) will runafoul of some versionof a feature-checkingtheory.A reviewer
adduces (i) as an example that plausibly meets all feature-checkingrequire-
ments, and for which we would thus need to appeal to the Chain Conditionor
something similar.
(i) *[ThatJohn believed mej to have appearedthat Maryis sincere] con-
vinced them that it was told ti.
The pronounme is merged into the passive VP with told, where it receives a
0-role but no Case. Next, a copy of me is mergedinto the 0-less subjectposition
of the raising predicateto have appeared, where it checks Case features via
exceptional Case marking.Finally, Merge introducesthe dummy pronounit
into the 0-less Case-checkingposition, [Spec, IP] of the passive was told. Note
that this does not invalidate the point in footnote 9-namely that, following
Chomsky (1995b), we take the C-CommandConditionon chains to be partof
the definition of Merge (Chomsky's Move a).

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions
354 SQUIBS AND DISCUSSION

(8) InterarborealXP-movementvia substitution(hypothetical)

X'

X DPi YP

DPi ,

. Y. ..
XP

YP X'

DPi Y, X DPi

. Y.. .

YP), in which neither c-commands the other. Thus, the operationis

illegitimate. The independentlymotivated Chain Condition (Chom-
sky's 1995b:253C-CommandCondition)is alone sufficientto exclude
interarborealoperationsof XP-substitution,and to significantly con-
straininterarborealoperationsinvolving adjunction.12
The second partof Rizzi's ChainCondition,(7b)-the "relativ-
ized minimality" condition-is discussed at length in Rizzi 1990.
Within the frameworkcurrentat the time, this was stated in terms of
government.We state it now solely in termsof c-command:the requi-
site relationshipbetween multiple instances of a single element must
be of minimal length (i.e., local). The notion "shortest" or "closest"
cannot be defined on interarborealoperationsper se; rather,it must
be defined on the links of the chains createdby such operations.Thus,
locality must be stated in terms of Relativized Minimality(7b) or the
Minimal Link Condition (Miyagawa 1993, Chomsky 1993). Again,
the definition of locality need not be a condition on final representa-

12
Note that certain interarborealXP-adjunctionstructureswould not run
afoul of the Chain Conditionin this way. For instance, if the DP in (8) were
to adjoin to YP-instead of substitutingfor the specifier of YP-then the DP
would c-commandthe other instance of DP in the complement-to-Xposition.
Again, a feature-checkingtheory may suffice to exclude the relevant deriva-
tions, although we leave the matteropen.
For an alternativeanalysis of interarborealXP-movement, see Nunes's
(I 995:chap.4, 1996) analysisof parasiticgaps andacross-the-boardextractions.
Although similar in spirit to ours, it differs in its assumptions about chain
formationand the Chain Condition.

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions
 SQUIBS AND DISCUSSION 355

tions, butmay be statedas partof Merge/Move,just as (7a), Chomsky's

C-CommandCondition, is so stated.

4 Conclusions
The theory developed in Chomsky 1993, 1995a,b does not exclude
the possibility of interarborealoperationsfor any principledreason.
Nevertheless, this possibility has remainedunexploredin the current
literature.If interarborealoperationsare not permittedby Universal
Grammar,then this is a gap in the theory that awaits a principled
explanation.We have suggested instead that this is not a flaw of the
theory.In particular,we have shownthatadmittinginterarborealopera-
tions provides a solution to the apparentlyproblematicnatureof head
movement for the ER. Furthermore,we have shown that the Chain
Conditionconstrainsinterarborealoperationsin such a way as to ex-
clude interarborealXP-substitution,thereby significantly restricting
the generative power of these operations. Furtherempirical conse-
quences of our proposals are left for futureresearch.

References
Baker, Mark. 1988. Incorporation:A theory of grammaticalfunction
changing. Chicago: University of Chicago Press.
Bobaljik,JonathanDavid. 1995a. In terms of Merge:Copy and head-
movement. In MIT workingpapers in linguistics 27: Papers
on minimalistsyntax, 41-64. MITWPL,Departmentof Lin-
guistics and Philosophy, MIT, Cambridge,Mass.
Bobaljik,JonathanDavid. 1995b.Morphosyntax:The syntaxof verbal
inflection. Doctoral dissertation,MIT, Cambridge,Mass.
Brody, Michael. 1995. Lexico-Logical Form: A radically minimalist
theory. Cambridge,Mass.: MIT Press.
Brown, Samuel. 1995. A possible account of cyclic head adjunction:
"Inter-arboreal"operations. Ms., HarvardUniversity, Cam-
bridge, Mass.
Chomsky, Noam. 1993. A minimalist programfor linguistic theory.
In The viewfrom Building20: Essays in linguisticsin honor of
SylvainBromberger,ed. KennethHale andSamuelJay Keyser,
1-52. Cambridge,Mass.: MIT Press.
Chomsky, Noam. 1995a. Bare phrase structure.In Governmentand
Binding Theoryand the MinimalistProgram, ed. GertWebel-
huth, 383-439. Oxford:Blackwell.
Chomsky,Noam. 1995b. Categoriesandtransformations.In TheMini-
malist Program, 219-394. Cambridge,Mass.: MIT Press.
Collins, Chris. 1995. Towarda theory of optimalderivations.In MIT
workingpapers in linguistics27: Papers on minimalistsyntax,
65-103. MITWPL,Departmentof LinguisticsandPhilosophy,
MIT, Cambridge,Mass.
Epstein, Samuel David. 1989. Adjunction and pronominalvariable
binding. LinguisticInquiry20:307-319.

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions
356 SQUIBS AND DISCUSSION

Epstein,Samuel David. 1994. Un-principledsyntax and the derivation

of syntactic relations. Paper presented at the Forum in Syn-
chronic Linguistic Theory, HarvardUniversity.
Frank,Robert,andAnthonyKroch.1995. Generalizedtransformations
and the theoryof grammar.StudiaLinguistica49(2):103-151.
Freidin,Robert.1994. Generativegrammar:The principlesandparam-
eters framework.In The encyclopediaof language and linguis-
tics, vol. 3, ed. R. E. Asher and J. M. Y. Simpson, 1370-1385.
Oxford:PergamonPress.
Groat, Erich, and John O'Neil. 1994. Spell-Out at the LF interface.
Ms., HarvardUniversity, Cambridge,Mass.
Kayne, Richard S. 1994. The antisymmetryof syntax. Cambridge,
Mass.: MIT Press.
Kitahara,Hisatsugu. 1994. Target a: A unified theory of structure
building. Doctoral dissertation, Harvard University, Cam-
bridge, Mass.
Kitahara,Hisatsugu. 1995. Target ot: Deducing strict cyclicity from
derivationaleconomy. LinguisticInquiry26:47-75.
Lasnik,Howard,andMamoruSaito. 1992. Move a. Cambridge,Mass.:
MIT Press.
May, Robert. 1985. Logical Form. Cambridge,Mass.: MIT Press.
Miyagawa, Shigeru. 1993. LF Case-checkingand MinimalLink Con-
dition. In MIT workingpapers in linguistics 19: Papers on
Case and agreementII, 213-254. MITWPL, Departmentof
Linguistics and Philosophy, MIT, Cambridge,Mass.
Nunes, Jairo. 1995. The copy theory of movement and linearization
of chains in the Minimalist Program. Doctoral dissertation,
University of Marylandat College Park.
Nunes, Jairo. 1996. On why traces cannot be phonetically realized.
In NELS 26, 211-226. GLSA, University of Massachusetts,
Amherst.
Pesetsky,David. In preparation.Optimalityprinciplesof sentencepro-
nunciation.MIT, Cambridge,Mass.
Rizzi, Luigi. 1986. On chain formation.In Syntaxand semantics 19:
The syntax of pronominal clitics, ed. Hagit Borer, 65-95. Or-
lando, Fla.: Academic Press.
Rizzi, Luigi. 1990. RelativizedMinimality.Cambridge,Mass.: MIT
Press.
Travis,Lisa deMena.1984. Parametersandeffects of wordordervaria-
tion. Doctoral dissertation,MIT, Cambridge,Mass.
Watanabe,Akira. 1995. Conceptualbasis of cyclicity. In MITworking
papers in linguistics 27: Papers on minimalist syntax,
269-291. MITWPL,Departmentof Linguistics and Philoso-
phy, MIT, Cambridge,Mass.

This content downloaded from 128.122.70.239 on Fri, 14 Feb 2014 09:49:49 AM

All use subject to JSTOR Terms and Conditions